Risk of Incident Circulatory Disease in Patients Treated for Differentiated Thyroid Carcinoma With No History of Cardiovascular Disease

Konstantinos A. Toulis; David Viola; George Gkoutos; Deepiksana Keerthy; Kristien Boelaert; Krishnarajah Nirantharakumar


Clin Endocrinol. 2019;91(2):323-330. 

In This Article

Materials and Methods

This was a population-based, open cohort study with prospectively collected data in which patients with DTC were compared to matched patients without DTC (controls) up to a 1:5 ratio. Age, sex, body mass index (BMI) and smoking status were the matching parameters. Data were derived from The Health Improvement Network database (THIN). THIN data are generated from longitudinal data documented in electronic medical records by general practitioners during each consultation using Read Codes, a hierarchical coding system for structured storage of information.[17] More than 675 practices, scattered representatively around the UK, contribute data to THIN covering 3.7 million active patients (6%-7% of the UK population).[18] THIN data are generalizable for the UK for major health conditions[19] and used for investigating circulatory disease outcomes and mortality.[20,21]

The study period extended from 1 January 1996 (study start or earliest index date) to 1 May 2016 (study end or latest index date). Individuals were eligible for inclusion in the exposed cohort (patients with thyroid cancer) if they were aged 18+ years at the index date, had a diagnosis of DTC and had no documented circulatory disease at or before their index date. Read Codes used to identify exposure are detailed in the Appendix S1. For each exposed patient (patient with thyroid cancer), up to five unexposed controls (individuals without thyroid cancer) were selected from patients registered in the same participating general practice. Controls were individually matched to cases on age at index date (to within 2 years), BMI (to within 2 kg/m2), smoking status and sex, provided that they were not diagnosed with thyroid cancer either before their index date or during the observation period.

The outcomes were ischaemic heart disease, stroke, heart failure and atrial fibrillation during the observation period. Read Codes used to identify exposure are detailed in the Appendix S1. Study participants were censored when an outcome was recorded.

Potential confounders, selected on the basis of biological plausibility, were used as model covariates in addition to the matching parameters. These were the presence of hypertension, diabetes mellitus, use of lipid-lowering medications and Townsend deprivation index (a measure of socioeconomic status[22] reflecting material deprivation). Information on Charlson's Comorbidity Index (CCI) was also extracted, and CCI was calculated and used as a model covariate[23] when all-cause mortality was estimated. This index, incorporating the comorbidities burden which may impact survival, shows marked predictive power for mortality[24] and is validated for use in primary care setting.[25] We modified the score to exclude cancer as this was the exposure of interest.

Statistical Analysis

Data extracted were analysed using appropriate descriptive statistics, with continuous variables described as mean/median (standard deviation/interquartile range) and categorical variables as proportions. A proportional hazards model using Cox regression analysis was used for the calculation of crude hazard ratios (HR) and adjusted hazard ratios (aHR), offset by the person-years of exposure to determine whether DTC was associated with an increase in the circulatory risk. Point estimates of the hazards ratios were calculated with 95% confidence intervals (CI), and a statistical significance threshold was set at 0.05. All analyses were performed in STATA 14.0 MP.

Sensitivity Analyses

A sensitivity analysis was undertaken by excluding those patients with prevalent disease (historical diagnosis of thyroid cancer, defined as diagnosed before study start) and their respective controls (thus, using only the subset of incident, newly diagnosed DTC cases and their respective controls). This analysis was selected since prevalent cases may represent cases with survival bias and in whom suppressive levothyroxine therapy may also be less likely. Finally, a sensitivity analysis focusing on those patients with documented TSH suppression (<0.1 mIU/mL) and their respective controls was also planned but not conducted because the majority of measurements were performed in a tertiary (hospital) setting and, therefore, were not available in the primary care database. Registration and authorization were obtained from the relevant Scientific Review Committee (16THIN092).